Structural safety assessment has evolved from the traditional allowable stress design to more rational methods based on limit states. Current practice on the limit states verifications are based on the partial factor method. However, since the beginning of this century, the arbitrariness in the definition of the partial factors used to increase the values of the loads and to reduce the properties of the material has been recognized. This simplified procedure are often conservative and may lead to expensive upgrades. This situation has motivated the development of methods for safety verification based on a probabilistic approach. Parallel to this, it is sought to develop, within deterministic methods, processes more compatible with real situations of rupture of structures, such as the global safety approach. Structural design standards are progressively introducing these new methods. However, a systematic analysis for the quantification of probabilistic safety resulting from the application of these methods has not yet been found in the literature. The objective of this thesis is to contribute in this safety assessment of reinforced concrete structures, analyzed by the conventional methods of Limit States, by Global Safety methods and by complete probabilistic analyses. Within the probabilistic analyses, it is tried to overcome a great difficulty in the application of the methods that is the evaluation of the probabilistic characteristics of the variables, by the application of a Bayesian approach, that allows a continuous revaluation of the variables, as well as the application in this definition of "engineering judgment" of the analyst. . An important point of the analysis that the reliability analysis usually done in isolated sections of a structure can lead to conservative and misleading results, since the structures behave as a whole.